THE BLOG
02/29/2016 05:54 pm ET Updated Dec 06, 2017

Peter Wittek, a roving adventurer between machine intelligence and quantum physics

Peter Wittek and I met more than a decade ago while he was an exchange student in Singapore. I consider him one of the most interesting people I've met and an inspiration to us all.

Currently, he is a research scientist working on quantum machine learning, an emergent field halfway between data science and quantum information processing. Peter also has a long history in machine learning on supercomputers and large-scale simulations of quantum systems. As a former digital nomad, Peter has been to over a hundred countries, he is currently based in Barcelona where, outside work hours, he focuses on dancing salsa, running long distances, and advising startups.

2016-02-28-1456691593-1100960-IMG_2684.jpg

S: Remind me again, what's your background?

P: Thanks, Sriram. I graduated with a masters in mathematics in Budapest, then I graduated from the National University of Singapore with a PhD in machine learning. Since then, I worked on a number of topics ranging from designing learning algorithms for massively parallel architectures to quantum physics simulations. Currently I am in a group of quantum information theory in ICFO-The Institute of Photonic Sciences in Barcelona and I am also affiliated with the University of Borås in Sweden. We work a lot on quantum non-locality, and when I have time, I work on my pet topic, which is quantum machine learning.

S: I've heard of machine learning. What is quantum machine learning?

P: It is an emergent field in the intersection of quantum information processing and artificial intelligence. Basically the fundamental question is this: can we build a smarter AI if we have quantum resources? This is, of course, a simplification, but my bet is that the next major application of quantum technologies will be machine learning. We have already seen some fascinating demonstrations on actual data sets. I believe that generalizing classical statistical learning theory to the quantum case will help us break some barriers that we have been struggling with for decades.

Most of the work in this field focuses on speedups that we potentially gain or on increased density in storing patterns. These are, of course, very important considerations, but I don't think we should stop here. I am interested in what limits quantum physics imposes on statistical learning theory: once we understand these, we can derive novel approaches in machine intelligence. Who knows, we might develop a strong AI based on semi-quantum systems well before we have a universal quantum computer. But now I am hand-waving.

S: You lost me.

P: Aren't you supposed to be listening while conducting this interview?

S: Okay. What's this about hand-waving? You've published an expensive book!

P: Yes I did publish a book, but let's not oversell it: the book is a monograph with little original work. I was interested in getting both an understanding and an overview of the various approaches, and the material was growing fast, so it became a book.

I have a background in mathematics and computer science, and most papers on the subject were written by physicists. Initially, it was an alien world: many tacit assumptions are never explicitly spelt out, the notation is different, and the terminology can be confusing. I was lucky because I visited two different quantum information groups while writing the book, and I could ask arbitrarily stupid questions. That helped a lot in getting my bearings right. Once I understood the foundations, it was easy to identify the main research directions in quantum machine learning. Given how niche the topic is, I am surprised to see how well the book sells.

S: How many copies did you sell?

P: I am not sure if Elsevier would be happy if I disclosed the number, but it is much more than my initial estimate of ten copies.

S: I've spent years trying to get you involved with startups. Heeded my advice?

P: Very much so! The Creative Destruction Lab, a business accelerator in the University of Toronto, contacted me about a year ago whether I was interested in becoming an advisor on machine learning in general, and on quantum machine learning in particular. Of course I was. I failed as an entrepreneur a decade ago, but I love to see how the whole startup economy works. The lab gave me the title "Chief of Quantum Machine Learning", which I love, although I understand it has a tongue-in-cheek overtone. Through this lab, I got to know more startups than ever, and one is actually making effort to understand and use quantum information processing for learning. Naturally, I jumped on board, and now I am officially a scientific advisor for this venture.

Then, on the side, I have an interest in a venture in Singapore called Fuzzie, a mobile gifting app. They are doing great: they just won the most promising startup award, then they did brisk business during Chinese New Year and Valentine's Day. There is nothing quantum here, apart from the founder's determination that will tunnel through every potential barrier.

S: Do you have any other side projects?

P: There are too many of them. If a problem is interesting, I can't stop myself from wasting my time on it. The topics range from Greek mythology through computational semantics to Bose-Einstein condensates. I tend to underestimate the difficulty first, I usually think a problem will take a week to solve. Then, driven by stubbornness, I work on it for months, until I get the results I am happy with.

S: I remember when we climbed Mt. Kilimanjaro together in 2008 and you arrived at the peak a long time after I did. You looked more pathetic than stubborn.

P: I always appreciate that reminder. I came a long way since then. Mountaineering has become a part of my life, although I don't do it as often as I wish. A year and a half ago I climbed Lenin Peak: that was 7,134 metres (23,406 ft). I don't think I want to go higher than that: we sat around in the high camps for the better part of three weeks as we were acclimatizing to the altitude. Otherwise, I love running: this year I hope to run a trail marathon.

S: Mountaineering and running? What about traveling? This is your chance to gloat.

P: I try to combine my travels with running: the Queen has been to a 112 countries, and I have been to 104. So I choose my running destinations in countries where I have not been yet. I must beat the monarch.

S: At one point, you went to Afghanistan to dance salsa...

P: That was a while ago and the destination was undoubtedly a misfit. I restarted dancing again in Spain. I am rustier than ever, but it is slowly coming back. There are free salsa classes and parties in Barcelona every single night, so I am spoilt for choices.

S: What's next?

P: Apart from beating the Queen? I want to have more independence in my work, and, honestly, I am a bit sick of traveling and moving around. So I put a lot of effort into grants and applications that would allow me to create my own group. The geographical uncertainty is initially attractive in academia, but after many years of traveling and frequent relocations, I wish I had a place where I knew I could stay for some time to come.